Research map / Vision-language and multimodal / CLIP and zero-shot
Negation-Aware Test-Time Adaptation for Vision-Language Models
Haochen Han, Alex Jinpeng Wang, Fangming Liu, Jun Zhu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add semantics reversion loss L_sr to entropy loss (ablation) | NEAT · MS-COCO MCQ-Neg | accuracy: 28.69 -> 45.91 | Improved |
| add textual debiasing loss L_tri to L_ent + L_sr (ablation) | NEAT · MS-COCO MCQ-Neg | accuracy: 45.91 -> 48.41 | Improved |
| add textual debiasing loss L_tri to L_ent + L_sr (ablation) | NEAT · MS-COCO Retrieval-Neg | I2T R@1: 44.96 -> 48.98 | Improved |
| add entropy loss L_ent to base CLIP (ablation) | NEAT · MS-COCO Retrieval-Neg | I2T R@1: 45.30 -> 48.06 | Improved |
| add entropy loss L_ent to base CLIP (ablation) | NEAT · MS-COCO MCQ-Neg | accuracy: 39.25 -> 28.69 | Hurt |
| add L_ent to L_sr + L_tri (ablation) | NEAT · MS-COCO Retrieval-Neg | I2T R@1: 7.12 -> 48.98 | Improved |
| use negation-separated prediction instead of KNN-based prediction (ablation) | NEAT · zero-shot top-1 prediction | accuracy | Improved |
| apply NEAT test-time adaptation to CLIP | CLIP · MS-COCO Retrieval-Neg | rSum: 326.47 -> 356.01 | Improved |
| apply NEAT test-time adaptation to NegCLIP | NegCLIP · MS-COCO Retrieval-Neg | rSum: 391.80 -> 402.55 | Improved |
| apply NEAT versus TCR online TTA | CLIP · MS-COCO MCQ-Neg | total accuracy: 27.71 -> 48.41 | Improved |
| apply TENT/SAR/TCR entropy-style TTA | CLIP · MS-COCO MCQ-Neg | total accuracy: 39.25 -> 26.43 | Hurt |
| apply READ TTA | CLIP · MS-COCO MCQ-Neg | total accuracy: 39.25 -> 39.26 | No clear effect |
Closest papers by meaning
- Vision-Language Models Do Not Understand Negation (CVPR 2025 · similarity 0.92)
- Frustratingly Easy Test-Time Adaptation of Vision-Language Models (NeurIPS 2024 · similarity 0.91)
- What "Not" to Detect: Negation-Aware VLMs via Structured Reasoning and Token Merging (ICLR 2026 · similarity 0.90)
- Revisiting the Role of Language Priors in Vision-Language Models (ICML 2024 · similarity 0.89)
- Realistic Test-Time Adaptation of Vision-Language Models (CVPR 2025 · similarity 0.89)
- Synthesize Diagnose and Optimize: Towards Fine-Grained Vision-Language Understanding (CVPR 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
